Enhancing integrated earthquake simulation with high performance computing
Introduction
While basic physical characteristics of the earthquake hazard have not changed, earthquake disaster evolves itself as society changes. In Japan, for instance, the earthquake disaster impacts on domestic and global economy activities will be tremendous if an inter-plate earthquake hits large cities. The shortage of the work force in the construction industry might cause difficulties in repairing damaged assets. To improve preparation, the first step will be to develop a hazard map of next generation which is able to show such consequences of the earthquake.
The authors have been developing integrated earthquake simulation (IES) as a candidate of a simulation-based prediction system for earthquake hazard and disaster [1], [2], [3]. This simulation is a seamless simulation of the following three processes: (1) the earthquake hazard process in which earthquake waves propagate in the crust; (2) the earthquake disaster process in which structures are shaken by strong ground motion; and (3) the anti-disaster action process which includes urgent evacuation or recovery of social functions. Simulation of the third process is essential in predicting secondary or indirect disasters which includes impacts on economic activities. Suitable models are constructed from Geographic Information System (GIS) and other data resources, and the prediction is made by analyzing these models with numerical analysis methods.
High performance computation (HPC) is required for IES, when it is used for a larger urban area. In particular, HPC is essential for the earthquake disaster process that makes use of linear or non-linear numerical analysis methods to compute responses of structures located in the area; strong ground motion is generated by earthquake waves which are synthesized in the simulation of the earthquake hazard process, and the seismic responses are computed for all structures by inputting the strong ground motion at a site of each structure.
In order to enhance IES with HPC, the authors are making the following two improvements: (1) layer-based design of IES architecture and the Common Modeling Data (CMD) class for inter-layer communication; and (2) the GenericDistributable class and the MPI_Process_Manager class for sending data of complicated structure and massive amount through Message Passing Interface (MPI) library. The basic concepts of these improvements are not original; the authors regard making such improvements as a standard practice to enhance an existing system with HPC. However, since IES uses data of various kinds, complex structure, and huge size, efforts have to made to materialize the concepts.
The contents of this paper are as follows. In Section 2, the first improvement that is related to the IES architecture is explained, and advantages of the current IES architecture are discussed. In Section 3, the second improvement for parallel computing is explained. Emphasized is the treatment of data of complex structure and huge amount through the MPI library. By using IES which is enhanced with HPC, a simulation of the earthquake disaster process is made for a part of Tokyo in which around 10,000 structures are located. The simulation results are presented in Section 4, and scalability of this IES computation is discussed. Concluding remarks are made in Section 5. It should be noted that IES employs C++, and a class is used to designate an object which is used in IES.
Section snippets
Drawbacks of wrapper-based design
Originally, IES employs wrapper-based design to facilitate data conversion among the three processes of the IES simulation as well as the visualization of the simulation results [2]. A wrapper is constructed for each simulation program of seismic structure analysis, so that it is responsible for interpreting input and output data for the analysis method. More specifically, the wrapper reads data from GIS and other data resources, and converts and combines the data to an appropriate form so that
GenericDistributable class and MPI_Process_Manager class
Among HPC architectures, clusters have the most popularity since they are more economical and accessible. In parallel computation, data have to be transferred among cluster nodes by using the MPI library. For IES, however, the use of the standard MPI library is not feasible [6], [7]. This is because CMD class has complicated data structure. Data stored in CMD cannot be transferred by using the MPI library as well as its extensions [7]. To overcome the inability for IES to use the MPI library,
Model construction
Two analysis methods for the seismic response analysis are used to measure the performance of IES that is enhanced with HPC. They are linear Multi-Degree-Of-Freedom (MDOF) analysis and non-linear Distinct Element Method (DEM) analysis; complexity of the analysis methods is examined by using these methods which require utterly different computational resources. For one building, two models are constructed from the data layer of GIS, one for linear MDOF and another for non-linear DEM. The data
Conclusion
This paper presents the two improvements of IES to enhance it with HPC by means of parallel computing. Major techniques developed are CMD and GenericDistributable classes; although the authors do not think that the basic concept of GenericDistributable class is original, well-structured CMD and GenericDistributable classes are developed so that complicated and huge data for structures are easily handled through several layers of IES. In particular, CMD class improves task distinction between
Acknowledgments
We are grateful for Associate Prof. Aiko Furukawa (Kyushu University) for providing the source code of the DEM analysis method, and Pasco Co. for providing GIS data. We are grateful for the support provided by the Japanese Government (MEXT).
References (10)
Introduction to computational earthquake engineering
(2006)- et al.
Improvement of IES wrapper for plug-in of advanced simulation programs
J Appl Mech
(2007) - Ichimura T, Hori M. Development of prototype of integrated earthquake disaster simulator using digital city and strong...
Software engineering, 8/E
(2006)- et al.
Data conversion of GIS to urban area model for natural disaster simulation
J Appl Mech
(2008)
Cited by (27)
A web-based system for the selection of characteristic ground motions
2019, Advances in Engineering SoftwareCitation Excerpt :The use of nonlinear seismic response analysis for the design and performance assessment of structures has become an acceptable alternative due to the increased computational power and parallel processing capabilities of modern computers which are now available to engineers [1]. It was also shown that advanced simulations of the nonlinear seismic response analysis of structures can be successfully performed by means of high-throughput computing (HTC) [2], together with parallel computing [3]. On the other hand, the low-cost high-performance computing systems which are available to most engineers are also able to solve most of the problems in this field, but they have to be used efficiently, and the mechanisms which are most suitable for any given application have to be identified [4].
A Prototype Seismic Loss Assessment Tool Using Integrated Earthquake Simulation
2018, International Journal of Disaster Risk ReductionCitation Excerpt :Then, nonlinear structural response analysis of all structural models is performed simultaneously to predict and illustrate structural damage of all buildings in the urban area. Since HPC is a key to carry out a large number of buildings, OpenMPI application has been applied to IES in order to enable parallel processing on the Central Processing Units (CPUs) [32,35,4]. Mostly, a statistical method using fragility curves developed based on historical data is used to predict building damage as a function of seismic intensity.
High Performance Computing for Regional Building Seismic Damage Simulation
2017, Procedia EngineeringBuilding seismic response and visualization using 3D urban polygonal modeling
2015, Automation in ConstructionCitation Excerpt :RemeshedExteriorPolys. IES is capable of determining the dynamic parameters of building numerical model, e.g., the mass, inter-story stiffness and inter-story strength, according to the plan of each floor [9,11]. When the floor plan abruptly changes along the height, these quantities of the corresponding floor determined by IES will be significantly different from the adjacent floors, thereby resulting in displacement concentration due to the vertically irregular floor plan layout.
A coarse-grained parallel approach for seismic damage simulations of urban areas based on refined models and GPU/CPU cooperative computing
2014, Advances in Engineering SoftwareCitation Excerpt :However, the computing platform is an important constraint of IES. A very powerful computer platform is necessary when the results must be obtained in a very limited time (e.g., for some emergency response purpose) [15]. For a large urban region, only supercomputer systems can meet the requirements, which demand a very high maintenance cost.